[PATCH net] r8169: fix powering up RTL8168h

[PATCH net] r8169: fix powering up RTL8168h

[Heiner Kallweit]

Since commit a92a08499b1f "r8169: improve runtime pm in general and
suspend unused ports" interfaces w/o link are runtime-suspended after
10s. On systems where drivers take longer to load this can lead to the
situation that the interface is runtime-suspended already when it's
initially brought up.
This shouldn't be a problem because rtl_open() resumes MAC/PHY.
However with at least one chip version the interface doesn't properly
come up, as reported here:
https://bugzilla.kernel.org/show_bug.cgi?id=199549

The vendor driver uses a delay to give certain chip versions some
time to resume before starting the PHY configuration. So let's do
the same. I don't know which chip versions may be affected,
therefore apply this delay always.

This patch was reported to fix the issue for RTL8168h.
I was able to reproduce the issue on an Asus H310I-Plus which also
uses a RTL8168h. Also in my case the patch fixed the issue.

Reported-by: Slava Kardakov <ojab@ojab.ru>
Tested-by: Slava Kardakov <ojab@ojab.ru>
Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
---
This patch will not apply to net-next as it conflicts with other
changes which have been done in the meantime. So I'll send a
separate patch for net-next.
---
 drivers/net/ethernet/realtek/r8169.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/drivers/net/ethernet/realtek/r8169.c b/drivers/net/ethernet/realtek/r8169.c
index 604ae783..c7aac1fc 100644
--- a/drivers/net/ethernet/realtek/r8169.c
+++ b/drivers/net/ethernet/realtek/r8169.c
@@ -4981,6 +4981,9 @@ static void rtl_pll_power_down(struct rtl8169_private *tp)
 static void rtl_pll_power_up(struct rtl8169_private *tp)
 {
 	rtl_generic_op(tp, tp->pll_power_ops.up);
+
+	/* give MAC/PHY some time to resume */
+	msleep(20);
 }
 
 static void rtl_init_pll_power_ops(struct rtl8169_private *tp)
-- 
2.17.0

Re: [PATCH net] r8169: fix powering up RTL8168h

[ojab //]

On Mon, May 7, 2018 at 7:11 PM, Heiner Kallweit <hkallweit1@gmail.com> wrote:
> Since commit a92a08499b1f "r8169: improve runtime pm in general and
> suspend unused ports" interfaces w/o link are runtime-suspended after
> 10s. On systems where drivers take longer to load this can lead to the
> situation that the interface is runtime-suspended already when it's
> initially brought up.
> This shouldn't be a problem because rtl_open() resumes MAC/PHY.
> However with at least one chip version the interface doesn't properly
> come up, as reported here:
> https://bugzilla.kernel.org/show_bug.cgi?id=199549
>
> The vendor driver uses a delay to give certain chip versions some
> time to resume before starting the PHY configuration. So let's do
> the same. I don't know which chip versions may be affected,
> therefore apply this delay always.
>
> This patch was reported to fix the issue for RTL8168h.
> I was able to reproduce the issue on an Asus H310I-Plus which also
> uses a RTL8168h. Also in my case the patch fixed the issue.
>
> Reported-by: Slava Kardakov <ojab@ojab.ru>
> Tested-by: Slava Kardakov <ojab@ojab.ru>
> Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>

Maybe also
Cc: stable@vger.kernel.org # v4.16+
?

> ---
> This patch will not apply to net-next as it conflicts with other
> changes which have been done in the meantime. So I'll send a
> separate patch for net-next.
> ---
>  drivers/net/ethernet/realtek/r8169.c | 3 +++
>  1 file changed, 3 insertions(+)
>
> diff --git a/drivers/net/ethernet/realtek/r8169.c b/drivers/net/ethernet/realtek/r8169.c
> index 604ae783..c7aac1fc 100644
> --- a/drivers/net/ethernet/realtek/r8169.c
> +++ b/drivers/net/ethernet/realtek/r8169.c
> @@ -4981,6 +4981,9 @@ static void rtl_pll_power_down(struct rtl8169_private *tp)
>  static void rtl_pll_power_up(struct rtl8169_private *tp)
>  {
>         rtl_generic_op(tp, tp->pll_power_ops.up);
> +
> +       /* give MAC/PHY some time to resume */
> +       msleep(20);
>  }
>
>  static void rtl_init_pll_power_ops(struct rtl8169_private *tp)
> --
> 2.17.0
>

Re: [PATCH net] r8169: fix powering up RTL8168h

[Andrew Lunn]

On Mon, May 07, 2018 at 07:20:53PM +0000, ojab // wrote:
> On Mon, May 7, 2018 at 7:11 PM, Heiner Kallweit <hkallweit1@gmail.com> wrote:
> > Since commit a92a08499b1f "r8169: improve runtime pm in general and
> > suspend unused ports" interfaces w/o link are runtime-suspended after
> > 10s. On systems where drivers take longer to load this can lead to the
> > situation that the interface is runtime-suspended already when it's
> > initially brought up.
> > This shouldn't be a problem because rtl_open() resumes MAC/PHY.
> > However with at least one chip version the interface doesn't properly
> > come up, as reported here:
> > https://bugzilla.kernel.org/show_bug.cgi?id=199549
> >
> > The vendor driver uses a delay to give certain chip versions some
> > time to resume before starting the PHY configuration. So let's do
> > the same. I don't know which chip versions may be affected,
> > therefore apply this delay always.
> >
> > This patch was reported to fix the issue for RTL8168h.
> > I was able to reproduce the issue on an Asus H310I-Plus which also
> > uses a RTL8168h. Also in my case the patch fixed the issue.
> >
> > Reported-by: Slava Kardakov <ojab@ojab.ru>
> > Tested-by: Slava Kardakov <ojab@ojab.ru>
> > Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
> 
> Maybe also
> Cc: stable@vger.kernel.org # v4.16+

No need. Heiner correctly marked this for net, not net-next, so David
with do what is needed for it to goto stable.

     Andrew

Re: [PATCH net] r8169: fix powering up RTL8168h

[David Miller]

From: Heiner Kallweit <hkallweit1@gmail.com>
Date: Mon, 7 May 2018 21:11:21 +0200

> Since commit a92a08499b1f "r8169: improve runtime pm in general and
> suspend unused ports" interfaces w/o link are runtime-suspended after
> 10s. On systems where drivers take longer to load this can lead to the
> situation that the interface is runtime-suspended already when it's
> initially brought up.
> This shouldn't be a problem because rtl_open() resumes MAC/PHY.
> However with at least one chip version the interface doesn't properly
> come up, as reported here:
> https://bugzilla.kernel.org/show_bug.cgi?id=199549
> 
> The vendor driver uses a delay to give certain chip versions some
> time to resume before starting the PHY configuration. So let's do
> the same. I don't know which chip versions may be affected,
> therefore apply this delay always.
> 
> This patch was reported to fix the issue for RTL8168h.
> I was able to reproduce the issue on an Asus H310I-Plus which also
> uses a RTL8168h. Also in my case the patch fixed the issue.
> 
> Reported-by: Slava Kardakov <ojab@ojab.ru>
> Tested-by: Slava Kardakov <ojab@ojab.ru>
> Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>

Applied and queued up for -stable.

> This patch will not apply to net-next as it conflicts with other
> changes which have been done in the meantime. So I'll send a
> separate patch for net-next.

That's fine, I'll deal with it when I next merge net into net-next.

Sending another copy of this patch for net-next is not the way to deal
with this.  Just make me aware of the impending complict and I will
resolve it when I see it.

Thanks.